ScriptGrain (@scriptgrain)

In depth: ScriptGrain: A Writing Tool That Measures Your Voice Instead of Guessing at It

Written from ScriptGrain's own profile data — claims, answers and testimonials — in a measured human voice via ScriptGrain · last written 2026-10-02

Ask an AI to "write like me" and it will oblige, in the sense that a mannequin obliges when you ask it to wear your jacket. The jacket fits. Nobody would mistake the mannequin for you.

That gap, between a described voice and a measured one, is the whole premise behind ScriptGrain, a product from Adapt Progress Evolve Limited, based in the United Kingdom. It launched officially on 20 August 2026. Its pitch, in its own words, is that it analyses samples of your existing writing across 45 measured reference points, builds a reusable voice profile, and generates content that reads as though you wrote it rather than as generic AI output.

What It Actually Does

Most "voice" features in writing tools rely on a prompt. You describe your style (friendly, punchy, a bit sarcastic) and hope the model's idea of "punchy" matches yours. Honestly, it rarely does.

ScriptGrain works the other way round. You upload samples of writing you've already done. The tool then runs a stylometric analysis, measuring the habits that make text recognisably yours across 45 reference points, and stores the result as a voice profile. That profile is reusable, so it doesn't need rebuilding for every piece, and drafts are generated against it rather than against a paragraph of adjectives.

The product's listed capabilities are stylometric voice analysis, reusable voice profiles, voice-match scoring, an API and MCP server for AI assistants, and a free AI-writing check. Five things, and they hang together more tightly than that list suggests.

The Scoring Step

This is the part that separates the approach from a fancier prompt. Every generated draft is scored for how closely it matches the source voice. The score is computed from measured feature deltas, which is to say the differences between the draft and the profile on the same reference points used to build it.

And when the score falls short? A polish loop revises the draft toward the voice. So the tool doesn't simply produce text and hand it over; it checks its own work against a number and has another go if the number disappoints.

Why does that matter? Because "reads like you" is usually a feeling, and feelings are hard to argue with or improve. A measured delta at least gives the revision something to aim at.

Who It Is For

The tool is aimed at people whose writing is part of how they're known: founders, makers, operators, agencies writing for several clients. The plan structure says as much. Paid tiers are organised by how many voice profiles you need, which is a fairly direct clue about who is expected to be using it.

• Writer, £12 a month, for one voice profile.
• Operator, £29 a month, for three.
• Studio, £99 a month, for ten.
• Agency, from £299 a month, with unlimited profiles, 10 seats and 1,500 credits a month included.

Annual billing saves about 20%. The profile count is the sensible axis, really: a solo writer needs one voice, while an agency juggling a dozen client accounts needs a dozen.

The Free End

Now, the free tier deserves a proper look, because it's where a sceptic would start. The Reader plan includes one voice profile and one lifetime writing analysis covering all 45 reference points, plus a free draft, and no card is required. Since its official launch on 20 August 2026, the idea has been that you can see your own measured profile before you pay anything.

That's a sensible bet. A profile is easier to believe when it's describing you.

Two further free tools sit outside the plans. One is a stylometric AI-writing check, which needs no account. The other is an in-browser writing style analysis. Neither requires a commitment, and both use the same measuring logic as the paid product, pointed at a narrower question.

API and MCP for AI Assistants

Since August 2026, every plan, free tier included, comes with a public API and an MCP server. The practical effect is that an AI assistant can build voice profiles, draft in the measured voice, and score voice match without anyone opening the ScriptGrain interface.

Worth sitting with for a moment: the same profile that a person builds by uploading samples can be called on by an assistant working elsewhere. The voice becomes a portable object rather than a setting buried in one app.

How It Differs From Prompting

The comparison people reach for is ChatGPT, so it's worth stating plainly. Asking a general chatbot to write in your voice means describing that voice in a prompt. ScriptGrain's answer is that the profile is measured from your actual writing, reusable across every draft, and checked by a score after each one.

None of that makes the output magic. It makes it accountable. A draft either lands close to the profile or it doesn't, and the tool says which. The mannequin, in other words, finally gets measured for the jacket.